Download historical stock data and test your trading techniques to maximize profits.
This repository hosts code to download historical financial data from Yahoo Finance. You can then utilize that data, or any other source of historical stock data to test your buying and selling techniques to optimize your strategy.
- Download historical stock data using
YahooDownload.py
- Run
backtest.py
with buy and sell conditions - Analyze output data located in the output directory
- -h, --help (Help). Show help information.
- -d DIRECTORY, --directory DIRECTORY. Directory containing csv files of stock data.
- -b BUY_CONDITIONS, --buy_conditions BUY_CONDITIONS. Conditions under which a stock should be purchased (already owned stocks will not be purchased again.
- -s SELL_CONDITIONS, --sell_conditions SELL_CONDITIONS. Conditions under which an owned stock should be sold.
- -cf CONDITION_FILE, --condition_file CONDITION_FILE. File containing buy and sell conditions.
- -df CSV_DATE_FORMAT, --csv_date_format CSV_DATE_FORMAT. Python string format for date. Ex. "%Y%m%d" for dates like 19900223.
- -oc OPEN_COLUMN, --open_column OPEN_COLUMN. Zero-indexed CSV column for the open price.
- -cc CLOSE_COLUMN, --close_column CLOSE_COLUMN. Zero-indexed CSV column for the close price.
- -dc DATE_COLUMN, --date_column DATE_COLUMN. Zero-indexed CSV column for the date.
Conditionals are not Python style. You can write a combination of conditions as such:
stock:decrease_rank <= 20 && stock:price <= 100
. Supported boolean operations are AND (&&), OR (||), and NOT (~).
Some of the variables support membership as part of the conditions. For example, if you only want to buy certain stocks you could write
the following condition: stock in {'AAPL', 'GOOG', 'GOOGL'}
which will be true only for stock symbols AAPL, GOOG, and GOOGL.
For numeric variables arithmetic can be performed. For example you could write a sell condition like this: stock:price > stock:buy_price*2
which will only sell once the stock price has doubled.
Arithmetic can also be used for dates. For example, if you want to sell after 3 months this sell condition will work:
date:today >= date:buy+[3*date:months]
Instead of parenthesis, use brackets ([, ]). Parenthesis only work for the boolean, not for math.
Some variables support array-type indexing for history. That means you can check the variable values for previous days. A query like
date:days_of_history >= 5 && stock:price:5 < 30
will select stocks whose price was less than 30 5 days ago. An index of 0 is today. Indexes greater than 0 go back in history. Indexes less than 0 are not supported because you cannot trade based on future information.
All array variable conditions must be combined with date:days_of_history
by '&&' like the example above because there is a time when
no history is available. The days_of_history condition must come first so that the array variable is not evaluated. If you neglect to include this, the program will crash.
- "stock" or "stock:symbol" - Is the stock symbol. Can be checked for equality or membership, not inequality or arithmetic.
- "stock:open_price" - The openning price for the stock on the current day. Supports array indexing.
- "stock:close_price" - The closing price for the stock on the current day. This is the price that will be used as the purchase price. Supports array indexing.
- "stock:price" - Same as close_price. Supports array indexing.
- "stock:buy_price" - The price paid to buy the stock. Only to be used for sell conditions, not buy conditions.
- "stock:owned" - Boolean, true when the stock is owned. To be used in future versions. Currenly it works, but is essentially pointless.
- "stock:increase_rank" - The rank of this stock in descending order, ordered by percent change in the current day. i.e. a stock with an increase_rank of 0 had the greatest increase by percent for that day. Supports array indexing.
- "stock:decrease_rank" - The rank of this stock in ascending order, ordered by percent change in the current day. i.e. a stock with a decrease_rank of 0 had the greatest decrease by percent for that day. Supports array indexing.
- "stock:change_percent" - The decimal percentage change for thist stock for today. Decimal percent means that 10% will be 0.1. Supports array indexing.
- "date" or "date:today" - Today's date in Unix epoch time. "date:today" is an array variable. If you want yesterday, you can enter "date:today:1".
- "date:days_of_history" - Number of days of previous data available for the stock, (i.e. days since the stock was listed). To be used with array variables. "date:days_of_history" is an array variable, if you want the days of history for two days ago, enter: "date:days_of_history:2".
- "date:buy" - The Unix epoch time that the stock was purchased (if it was). Only to be used for sell conditions, not buy conditions.
- "date:day_of_week" - Today's day of the week as an integer from 1 to 7, where 1 is Monday and 7 is Sunday. "date:day_of_week" is an array variable.
- "date:month" - The month that today is in. 1 is January. "date:month" is an array variable.
- "date:days" - Constant value to be used in arithmetic. If you want to sell after 10 days the sell condition would be
date >= date:buy+[10*date:days]
. - "date:months" - Constant value to be used in arithmetic. Similar to "date:days".
- "date:years" - Constant value to be used in arithmetic. Similar to "date:months" and "date:days".
stock:decrease_rank < 5
stock:increase_rank < 5
date:date_of_history >= 1 && stock:change_percent:1 < 0 && stock:change_percent < 0
stock in {'AAPL', 'GOOG', 'SPX', 'CSX'}
stock in {'AAPL', 'GOOG', 'SPX', 'CSX'} && (stock:change_percent > 0.05 || stock:change_percent < 0)
date >= date:buy+[3*date:months]
(stock:price > stock:buy_price) && (stock:increase_rank < stock:increase_rank:1)
- This backtester does not currently support intraday data. This means that it only makes a trade (buy or sell) at the end of the day.
- No support for splits. All data provided to the backtester should be relative to the first day or last day. Yahoo Finance data does do this automatically.
This software is built in part using Booleano by Gustavo Narea licensed under the MIT License, a copy of which is in the file LICENSE. Booleano on launchpad. All changes I made to Booleano are available in the Booleano subdirectory of this repository. Changes made to Booleano were to make it compatible with Python 3, add arithmetic, and add array variables.
The rest of the code, all code not in the Booleano directory, was written by Michael Dombrowski and is licensed under the MIT License located at the top of all source files and in the LICENSE file.